Skip to content
How to ThinkIn the Age of AI

Periodically perform automated steps manually — maintain…

Periodically perform automated steps manually as calibration exercises to maintain intervention skill and detect when automation has drifted or is producing incorrect outputs.

When This Fires

  • When you have automated steps that you haven't performed manually in months
  • When automation fails and you realize you've forgotten how to do the step by hand
  • When automated output "seems off" but you can't tell because you've lost calibration for what "right" looks like
  • Complements Sovereignty check before automation — if the automation… (sovereignty check) with the ongoing skill-maintenance practice

Common Failure Mode

Complete delegation: "The automation handles it, I don't need to know how it works anymore." This is fine until the automation breaks, produces subtly wrong output, or needs modification. At that point, you've lost both the ability to intervene and the knowledge to evaluate what's wrong. The failure compounds because you may not even recognize the automation has drifted.

The Protocol

(1) For each automated step, set a manual calibration cadence: monthly for critical automation, quarterly for routine automation. (2) During calibration: perform the step entirely by hand, from inputs to outputs. (3) Compare your manual output to the automation's most recent output. Discrepancies indicate either automation drift (fix the automation) or your own skill drift (note what you'd forgotten). (4) During manual execution, note anything confusing, unclear, or surprising — these are knowledge gaps that would block you during an actual automation failure. (5) Update your automation documentation with anything you discovered during the manual run. If you needed information that wasn't in the docs, add it.